 My name is Angel and I'm going to be talking this morning. We have two actually quite similar presentations. You can see those as just one. And actually I'm going to be talking a bit about weather types and atmospheric circulation regimes with a bit more detail on the second presentation. But this one, I'm going to be talking about this intracisional impacts of the extra tropics on extreme rainfall events. And this is related to some work that we have been doing in southeastern South America and also northeastern US, northeastern North America and a few other places. But I'm going to try to focus on these two places just to illustrate a couple of ideas, especially in terms of extreme rainfall events. So the plan of this first talk is just to remind you or to discuss a few ideas about these events. And then we're going to talk about some of these typical mechanisms, large-scale and regional-scale mechanisms. But I'm sure that you have discussed last week in far more detail than I'm going to be discussing. And then we're going to go to this concrete example for southeastern South America and for the northeast US, in particular for the Ohio River basin, with some work with our team. You're going to see some work that Andy Robertson, who was here last week, was doing, I think, a couple of years ago. And then we're going to summarize that. And then we're going to explain how this weather typing thing works and if we could use it or not for this kind of research on tropical-extra-tropical interactions. So, well, I think that you know that extreme events is obviously something, but at least for me, super interesting because it's very impactful for society. And we have been having several of these extreme rainfall events, not only temperature extreme, but also rainfall events in different places of the Americas. In particular, we had a couple of really bad events in southeastern South America in Paraguay. And I'm going to visit that one. It's a paper that we're about to submit. But something really interesting to me, too, is that it's extremely difficult to fork as extreme events. So, you know, that's an interesting challenge. And it's related to the spatial and temporal characteristics of these events. And in general, there are, like, different physical mechanisms associated with these rainfall cases. And you have, of course, the massive-scale convective complexes. And since we're trying to focus a bit on this extra-tropical region, you also have these baroclinic fronts and, of course, heat and moisture conduction, for example, by the effect of low-level jets. And we have several of those in South America, you know, but also in North America. And then you have atmospheric rivers. So, in order to address all these different drivers or physical mechanisms, there are different ways. There are different approaches. But what we are trying to do here is trying to see if we can use this idea of synaptic control that probably you already know to try to understand, at least, or try to unify a few of these different physical mechanisms. I'm probably you, you know, are aware of these, you know, expressions that I'm writing over there. I think that Gorman's team had been, like, you know, publishing a lot about that. And the general idea, if you are working with extreme rainfall events that are, for example, associated with convective precipitation, you have a few ingredients that you require there. You need some initial lifting force, you know, for your parcel, and you need some moisture for sure. You need some instability. And the idea is you may have a few of those or all of them, and you might still not have extreme rainfall events. So what we want is to try to find what are those suitable conditions, what are those, even like a synaptic scale, what are those conditions that are conducive to the occurrence of these extreme rainfall events. And we like to use for that approach this weather typing method that we're going to be discussing a bit better later. I just want to show you this particular case for Southeast and South America, which is the box that you can see over there. And this is just a particular way to define an extreme rainfall index, which is non-local. You are basically considering all the grid boxes for either region of interest. There are, you know, your F, Y, J over there could be defined in terms of different thresholds, different percentiles. So you can have the 95th percentile, the 99th, or even like you can use that same definition to address, to analyze dry days. And so as I said before, one of the main problems when you are trying to forecast these extreme events is that it's very difficult to do that at a very local level. So what you do, you know, you are better at forecasting these events when you are actually considering like a big region. And it will be very nice to have kind of a more localized thing, especially, you know, there are a lot of things that we still don't understand that most of our seasonal and actually sub-seasonal forecast experiment are addressing these average rainfall events instead of these ones that, again, society considers extremely important. So what you have there in the blue line, which is the same for those three, is the average of rainfall for that box. And in the bars, you actually have, let's say, a number, a frequency of this extreme event for each one of those thresholds. And what I did, the bottom is just to indicate which of those years are Almenia years, which is EN, and then you have it strong with an S moderate or weak. And also I'm indicating which ones are neutral and which ones are, well, not all of them, and which ones are Almenia. So probably you remember that for this particular region, when you have an Almenia, you have more rainfall. And the probability of having extreme rainfall events is higher. And when you have a Almenia, it's just the opposite. But then, you know, a simple analysis there will show you that this idea of just using Almenia to forecast, which is a common thing in this region, just using a seasonal driver like Almenia to explain what's going to happen or what's going on in terms of rainfall, average rainfall, and also extreme events, it's not enough. You can see some cases in which you can actually have, like for example, 2010, you have an Almenia, which is moderate, but you also have a lot of dry days. So you have this combination of stuff, this combination of extremes happening at the same time. So several dry days and also a lot of days with a lot of rain. And what we are going to say, what we have been explaining, you know, about this kind of event, is you cannot use only this typical seasonal driver, like Almenia, to explain what's happening. And I can talk far more about that. But let me just go back to a few ideas that I'm sure that you discussed last week. And one of these is this typical concept of marginally propagating rosby waves. And what I'm showing there is a very simple experiment with the shallow water barotropic model. And you can see that this, you know, with a small perturbation close to the equator, under certain conditions, you can develop this kind of, I don't know if this has, you can develop this kind of wave that we actually, I tried to show you yesterday, but it leads to some of you and one of the labs. So the idea is, well, as you can imagine, that there is a connection between obviously the tropics and the esotropics, and in some cases that can actually go back again to the tropics. And what we want to focus on here is that if we analyze changes in the circulation pattern, we might be able to say something about the occurrence of these extreme rainfall events, as I explained before. Sorry, this is December, January, and February. Yeah, I mentioned to, I forgot to mention that. So what we saw here was also December, January, and February, and this is a period that I'm considering 1980 to 2010. Yeah, I'm going to change that a bit later. Okay, so I think I'm going to stay over here because I need to be closer with the computer. So we have seen this also in this workshop. So the idea is we can also have a link, a connection between the tropics and the esotropics, for example, impacting things like the NAO, the North Atlantic Oscillation. And this famous paper, Kasu, already showed that. And it's consistent with this idea of trying to understand what's going on in terms of the spatial and temporal characteristics of the circulation types, of the circulation regions. And as, you know, as there in the literature too, those waves can interact through different mechanisms with other subsistional and seasonal drivers in the region of interest in particular. We have, and I think that Mariano talked about this last Friday, but this is what's called the South Atlantic Conversion Zone. And it's associated with, which is, you know, subsistional phenomenon, subsistional driver if you want to call it that way. And it's associated with this South American level jet. And actually it's not only one South American level jet as we will see, or you know, probably Mariano already mentioned that too. And the idea is like there are a lot of these different complex interactions at multiple spatial and temporal scale happening in the region, the same region, but also with this tropical and exotropical interference if you want. So this is, we understand, you know, it's already in the literature. This is also associated with this other phenomenon like the low level jets as I was discussing before. And so I'm coming to this point, the idea of synaptic control, the idea of if I pay attention, if I try to really characterize and understand what's going on in terms of the atmospheric circulation and how those circulation patterns change in terms of location or frequency of occurrence or persistence, intensity, et cetera, we can try to understand better what's going to happen or what's happening with extreme rainfall events. So something that you can do, and it's not important right now how we define this, is to take a look at those typical or recurrent circulation types, circulation regimes, and you will be able to identify there a few of these baroclean propagation systems and also waves, you know, these other marginally propagating rosby waves that could start like, you know, close to the tropic and then they interact with the jet stream. And this is like obviously having a third view, but once you go to the region, you can recognize things that the forecasters at weather time scales see, let's say, every day. And the way you define these ones, these circulation types, like you can only have one of those per day. There are different ways to do that, but in this particular case, that's how I'm defining those. And you can see, in some cases, you can see directly like the low-level jet, the South American low-level jet, but as you probably know, there are different flavors of that one, the CHAC one, the no-CHACO event. And you can relate those, you know, the beauty of the approach, I think, is that you can explain what is the physical outcome, in this case, the rainfall extreme events or a sunny day or, you know, snowy day in terms of, well, not for December, January or February, but in general, in terms of links to the climate drivers, like ANSO, but also SACs, let's say, different time scales, different temporal scales, but also coming from the tropics or the exotropics, and the actual mechanism, like, for example, baroclinic fronts that may be impacting the region, that may be producing this event. And what do you do? Well, if you want to characterize, as I just said, like the relationship between the circulation patterns and the occurrence of another extreme event, you can do a few things. Like, the first one is just a composite analysis for each one of those regimes, which are built on a daily basis. You say, well, what is the typical rainfall pattern or what is the typical rainfall anomaly by composite of rainfall anomalies associated with each one of those, each one of the... Let me try again. Yeah, you fixed it before, but... Okay. Okay, so what I was saying is just, like, there are different ways to do this and this is just a composite analysis. And by the way, all these things that you are seeing here, we can reproduce and we are actually going to be reproducing a few of those in the labs. So if you think that this approach makes sense for your projects, just let us know, because we can tailor this afternoon for your projects, especially I'm going to be showing in the next presentation the things that I have planned for this afternoon. Okay, so... Well, you can see that if you're interested in extreme events, and again, this is just a composite, this is the average, but if you are interested in rainfall and a lot of rainfall or above normal rainfall, you should be paid attention to the circulation regimes that we're calling here D, E, and F. Okay, we'll go to the weather typing thing later. And... something else that you can do is just to do, you know, you can use physics and also on a statistical analysis, like in this case, and try to see which particular circulation regimes are associated with the occurrence of this extreme for each one of those thresholds that I mentioned before. And you can see that there is, for the regime number four and number six, maybe I needed to do... to have the computer. So, for these two, this one, you know, and you can see the low-level jet over there, and for this other one, you can see that those are statistically significant, you know, 95% of the P, sorry, 0.05. These two are statistically significant associated with the occurrence of those extreme events. But even when the five is not really statistically significant, because of the transitions that occur between the different patterns, circulation patterns, number five is actually also, you know, associated with extreme rainfall events. And for the dry events, actually one, two, and three are, you know, more commonly associated with those dry events that I mentioned before. So, I don't know if you are aware... this microphone is confusing me a bit, but I don't know if you are aware of this particular event that happened between December and February 2015-2016. It was reported as, you know, one of the worst things that occur in the actually lower Paraguay river basin in the last 18 years, and something like 130,000 people, you know, were displaced, and it was a really bad thing over there, and they weren't expecting that. The interesting thing is that the seasonal forecast actually suggested that something like that obviously was a seasonal forecast, so it was like three months, or something like that, and then we could actually have forecast that using the S2S database. But the important thing is if we want to understand what happened, we can also use like this synoptic control approach, this weather typing approach to identify what were the drivers. And once you do that, actually the answer is kind of, in the academic literature. But anyway, so what you can see on the lower panel there is the typical behavior for PSI, for the stream function. And then on the bottom panel, what you have is a series of this rainfall anomalies for one of those months. And in brief, what you can do there is just to analyze what happened. The question that we are trying to answer in this paper is to see if that was really an extreme event or it was a common extreme event, it was something that we could have forecasted well, but also it was like a traditional mechanism, and actually it was. And as you will see in a bit, it's just a no-chaco thing. And we can get to that conclusion using a traditional, if you want, EUF analysis, but also this kind of weather regime approach for different variables in this particular case for the circulation, for the stream function. And you will be able to, you know, you can identify, and that's the kind of stuff that we want to do in the labs too, which are the types of events. And in brief, I don't want to bore you with all the details, but in brief, there was a constructive interference between El Nino for that year, which as you might remember was one of the strongest on record. And then an initially high occurrence and persistence, let's call it like that, of the phase four of MJO that as you might have seen, we were playing with this yesterday in the lab, anything last week again with Andy. When you analyze the composite, when you do a composite analysis for the MJO phase four for that region, you can see that it has typical behavior of above normal, a lot of rainfall in the south of Paraguay, northern Uruguay and southern Brazil. And that's basically what you observed for that year. So this combination, this tropical-extra-tropical interaction, this combination, if you want cross-time scaling interference provided suitable conditions for the alternation of two of those weather regimes that I mentioned before, and maybe the most important, not maybe for sure, the most important was like this low-level jet, which is characterized as called a non-Chaco one. The Chaco one will produce a lot of rainfall in this region, northern Argentina, but the non-Chaco one actually tends to bring a lot of rainfall in that direction. And we know that there is a relation, every time there is a linear, there are enhanced probabilities of more advection of moisture from the Amazon, and that we also know that with MJO this kind of behavior tends to bring a lot of rain. So we could have said something, we could have forecast that event actually well in advance, at least two weeks, and probably a month in advance with the difference tool we had, and with the understanding of the physical mechanisms, and we could have saved so many lives actually. The same idea could be applied now to different but kind of similar events that is the exotropics too, and now what we're going to do is just to take a look using the same kind of approach to the northeast of the U.S., and all previous studies like Nakamura at all analyzed what were the conditions, the first circulation patterns associated with the occurrence of that kind of events, and this is the typical behavior, or this is the typical circulation pattern that you find associated with extreme rainfall on the Ohio River Basin, and it's associated as you can see with basically advection of heat and moisture from the Gulf of Mexico towards the region of interest, and you know has been like that in this particular analysis more than 100 years was analyzed and it's basically the same pattern it hasn't changed a lot, so if we see if we can forecast in advance this circulation type we might be able again to take advantage of the situation, and as you might know for dynamical model is far easier to forecast winds and advection and this kind of circulation types than to actually forecast rainfall or even extreme rainfall events as I mentioned before that's kind of complicated so the idea is try to take advantage of that situation, so this particular analysis that Andy performed was published in 2015, they extensively analyzed what are the conditions associated with the circulation type which is the one that Nakamura reported before too, and that actually we later analyze in other terms, but you can do the same kind of approach that I mentioned before which is now that I have now that I know which are the circulation regimes associated for that particular season and by the way in this case is MAM you can again use physics to help yourself with a bit of statistics to identify which ones are the weather regimes associated with extreme rainfall events and those are the ones that Andy and his team reported for that particular case number two and number four and you can see the number four is particularly important for extreme rainfall event and it's the one that has been reported before the other one which is a bit more of a coastal associated with coastal extreme rainfall events or even coastal rainfall events and something else that we have been doing is trying to analyze how well that is represented by different models and by different configurations in those models again trying to understand those tropical exotropical interaction and those cross time scale interference that I mentioned before for that particular regime models tend to be very good you know tends to be at least decent enough in terms of the representation of those events that couldn't that cannot be said for all the other cases all the other regimes but this is like good news if we are talking about extreme rainfall events for the northeast particularly you can see in this case it's again like the composites for the rainfall for that particular pattern, atmospheric circulation pattern and the different experiments run in particular with these GFDL models at different spatial resolutions provide a good representation of at least the spatial pattern of rainfall not necessarily the spatial characteristics of the of rainfall but it's good enough just to summarize these ideas for the first part of the talk I just wanted to say that of course there are different mechanisms and some of those might be related and you can say that there is some interaction between them but that might not be the case and the idea is can we have an approach based on synoptic or larger scale or regional scale atmospheric circulation patterns that can provide a good understanding of what's going on or what might happen and also that can help us identify what are the climate drivers associated with that kind of events like in this case extreme rainfall events in the exotropics and can talk a bit can say something about those interactions between the tropics and the exotropics between the different time scales between the different spatial scales and we think that indeed that's possible for example if we use the weather typing approach that several of people here in this room have been using and have far more experience than me but let me just go to the other presentation because something that we're going to be doing in the labs is I'm going to provide these codes and we're going to be able to produce those figures and more so those are tools that if you can see that are useful for your experiments you can basically assimilate right now and actually I'm going to be showing a particular example here for what we just did was MAM in the northeast of US and just we're going to choose basically the same region different season just to illustrate that the approach tends to be it doesn't work always but in this case it's robust enough as to work in this case for December, January and February and what we're going to be able to do is to build we can call it that love triangle between the synoptic or the circulation pattern the climate driver and the actual impact event that we are interested in like the extreme rainfall event in this case we don't need that to be an extreme rainfall one and something else that I'm going to be doing here is trying to explore the question of if we can take advantage of those top tropical exotropical interactions to have enhance improve a scale not only for seasonal scale but also a scale and this has to do with a few I'm going to be like going back to a couple of papers that we have published on that topic and I don't know how familiar you are with this idea but you can you can try to always go to better resolution, higher resolution in your dynamical models or to have better statistical models or to have like better combination of those models let's call them hybrid models but you can also try to use something that in the literature sometimes is often called a coarse grain approach let's say that we want to understand what are the physical available states of the system so I cannot I cannot expect to see a transition regime with a happy face in the atmosphere there are obviously physical laws that are telling me these are the different states that the system can use and if I can characterize those states and I can say well the probability of the transition of my system from this particular state to this other one is I don't know this number and I know that this particular state with extreme rainfall events or sunny days or not I might have an additional tool I might have like additional information to say something about what might happen or not for those events that are of interest to be so as you can imagine that is often called a nonlinear dynamical perspective for to analyze this kind of system and a lot of people have been working on that in this particular case just like I like this figure I like to use this figure if you are not too used to the idea of this available set of the system and what team Palmer is basically telling us there is that those cups are your two options for your system and the other one is it doesn't rain and we might not have enough information to say exactly how the system is choosing between those two and how that is reacting to the effect of an external forcing but we might be able to use that nonlinear dynamical perspective or approach to try to provide additional information on what are the chances that it's going to rain or it's not going to rain you know one way to understand that approach you can imagine that you have for a particular season or for a particular decade you have let's say three available states A, B and C and if you think in terms of that approach you might say whatever happens might be understood as a combination as a sequence of those available states so for example a sunny day might be just A, A, A, A or maybe an extreme rainfall event like the ones that I just showed in the previous presentation might be something like A, B, C or A, B, B, A so if you identify you can there are some similarities here between the circulation regimes and circulation types and an alphabet that are the words associated with the different available states but also what are the letters sorry but also what are the words that are associated with your particular event of interest and you will be able to write that down every time I see cold front happening over this region that tends to be associated with something like B, B, A or B, B, C for example grain approach you might be able to use these circulation times and the sequences of circulation types to say something about the system and this brings us to the idea of how to define those available states and that is not a trivial problem that's actually a very complex one one way to approach one approach for those available states might be to use cluster analysis and there are different ways to do that but in particular we can use and this is what I have prepared for the labs in this afternoon you can use a K-means approach and basically what you are doing is let's say that you have for a typical three month season like 90 days and you have 30 years or so you have almost 3000 days and what you want to do is to classify 3000 days in a certain number of clusters and what you do is at the end you get you can use some statistics to get some help in terms of what is their adequate number what is a nice number of clusters to use but I think that the best is to use physics and actually to understand what are the physical mechanisms and see if those clusters that you are obtaining using your let's call it a blind method or something that you can recognize in terms of the physical processes for that region so let's say in this particular case there are seven clusters so these are similar days and these are the different circulation patterns that I have been talking about and again there are different ways to define this and actually for each method there are subtleties there are different ways to define different approaches but basically what you want in this case means approaches to minimize those distances that are typically associated or typically chosen as the Euclidean metric but you can use something else like a Mahalanobis Mahalanobis metric too and then there are different again ways to know to assess how well that classification method works for me and a typical classificability index used is that of Michelangeli at all from 1995 but again I think that to be able to recognize to be able to see if those particular clusters those weather types make sense or not from a physical point of view and if I can for example since those weather types are actually associated I can compute as I said before the probability of having a transition from weather type number one to weather type number three if I can see in some of those cases even like some ways propagating or if I can actually see that it makes sense the physical mechanism that those sequences are explaining so then there is a physical background there are physical reasons to decide how to approach that problem and then in this kind of approach we want to pay special attention to those daily transitions as I just mentioned the duration the persistence or not of those weather regimes and then the frequency characteristics at different time scale from subsistential to decadal even if you want to climate change those statistics provide a nice tool to understand what's going on not only in terms of those extreme events we also pay attention to the spatial patterns and we can use this method as I said before to identify candidate predictors for a model or how well we can even use this to diagnose how well my dynamical models are working for that particular region and season so there are several advantages of this approach so I'm going to show you what you are going to be running as the first experiment your lab this afternoon and again this is going to be for this region that I mentioned before in North America the North East North America a bit of Canada and the US and in this particular case is going to be December January and February and what we are going to be showing here is the different products or the different plots and the different tools that you can use that you might be able to use or not for your project of interest originally when we were discussing this with Fred and David and other people we just thought that we will have some exercises so people can work on different regions of the world with this same approach so for example South East and South America like the weather types that I showed before those typical six that have been reported by several authors then like the typical four for the North Atlantic European region and then something else for Africa and for Asia but this morning I was just discussing with Fred the possibility that we might just like focus on the actual project that you have the values of what we've done the V8 attention from those so it's up to you we will discuss that later so what we will be able to do is to you're going to have a MATLAB code we have this in different languages but we decided to use MATLAB for this lab and you're going to be running that and you're going to be able to get those five clusters, five circulation types again this December, January and February and on the top panel you are seeing the entire hemisphere but if you want you can change the region that you are using to define the weather regimes and then obviously one thing is that region that you use for your cluster analysis another different one is like the one that you use to visualize but in this particular case we got this five usually for December, January and February we tend to use four or five of these solutions we can talk more about that later and you will be able to have you know for your experiment the hemispheric view of those weather types and some of those you can even like recognize a few of those borrow cleaning like wave propagating, like wave trains that we have also mentioned before and then you will be able to do a zoom in that you can tailor, that you can customize if you want and in this particular case we are using geopotential height of 500 millibars but you can also use a different variable that's what you want, like what we did with this recent paper for the Paraguay river basin that we were just using as stream function so we will be able to help you with that if you want so let's say that we will pay attention to what are the special characteristics of those circulation patterns and those winds those vectors might just mean winds but something that could be nice to do is just the integral of the temperature transport along the column the atmospheric column to see in this case if I can have extreme rainfall events associated with the weather type number 3 which is similar to what I showed before in the MAM in March, April, May also it will provide you this composite for average rainfall that if you want to take a look at the extremes and also those other figures that I showed before when you are analyzing which of those weather types associated with that extreme events and you can define that threshold or whatever you call extreme so something else that I mentioned is that we also want to pay attention to the temporal characteristics for example the daily transition so I know that this plot looks like noise or even like art and actually it looks like I don't know if you are familiar with this clay the clay work so actually we are calling those things in 2015 clay diagrams and what you are seeing there is for in this case we have December, January and February and you have a certain number of yen in 1981 to 2010 so for each one of these columns you have let's say it's not a year it's basically a season December, January and February so your first tile over here is December 1st and your last one is the last day of your season which is in this case the last day of February and what you are seeing here is the daily evolution of those weather types so I don't know if this is really the clay diagram by itself it's useful but you can use it to build a few useful things in terms of characterizing or analyzing the temporal evolution of these patterns for example you can from that matrix from that clay diagram you can build a transition matrix and then you can have an idea of typically you can consider all those years or you can just have that for a subset for a smaller sample you can say well if I am today in weather type number one what's the probability that it persists and you can see that it's fairly high or that it actually goes to a different weather type it transitions to a different weather type which means that you can actually build some kind of a Markovian model formally a Markovian model with this kind of information that you are observing and you can also if you want to diagnose how well your dynamical models are reproducing these characteristics the temporal characteristic of your circulation regimes you can use also the transition matrix to diagnose to evaluate to assess if that's something that you are interested in of course then you can go to different time scales and you can just do some filtering that is suitable for the intracisional time scale and you can see how for your particular season December January and February how is the subsisional evolution or some people are calling this the subsisonality associated with each one what is the subsison evolution of each one of those types maybe some of them are going to be appearing more often at the beginning of the season or at the end there is some useful information that you can obtain like this particular weather type tends to appear more often at the very beginning of the season and it's associated with beautiful sunny days but these other two are associated with rainfall or are just like the precursors for a transition to rainy season that is going to come after this one and I can see them happening a lot at the end anyway so all these plots are going to be available you know for you in that code if you want to play a bit with that for your project and of course you can go to the next and the next and the next time scale so in this kind of plot you're going to be able to see the intracisional evolution of the frequencies of occurrence of your weather types and of course we only have 30 years in this case so we are not going to be able to say a lot about Decadal and definitely nothing about climate change but if you have enough years you might be able to use the same approach to go to those other time scales so this is what I think is interesting and this is what we did for those paper for that work that I started to discuss in the previous presentation but I didn't want to include this there just that discussion about the importance of using circulation regimes or atmospheric circulation patterns but this is one of the nice things about the method because then you can remember that link, that triangle that I mentioned before between once I have identified what are those proxies of the available state of the system I can try to identify what are suitable candidate predictors or climate drivers associated with them so you know there are different ways to do this but you can say well what are the typical SST patterns associated with each one of those weather types or if I have a really high frequency of occurrence of this particular weather type what's the typical SST pattern associated with that for example 80th percentile of occurrence of the frequency of that weather type so you can see well it's associated with an El Nino SST pattern or something that is of interest in the Atlantic or in the Indian Ocean so you can start playing with that and this is a particular example for SST but of course you can do that with different fields you can do the traditional approach trying to compute correlation again that's not with cause and effect but you can try to find some statistical correlations between the frequency of occurrence of your weather types you know different time scales in this case it's seasonal and for example things like El Nino as I just mentioned but also PNA or NAO and you know depending on your region you might have like more or less of these climate drivers so you might be able to say well you know I need to take a look at the statistical point of view every time I have in this case it's an El Nino but if this is like statistically significant over here it will be like El Nino every time I have El Nino I might have like some inhibition of this particular mechanism that is going to make you know that I have like a lot of more dry days in my season for this particular region or since I have a more El Nino so then probably I'm going to have more propagating rustic waves come in my direction and those are going to be affecting the circulation patterns that are going to be able to produce or not these extreme rainfall events in my region and also it's going to impact the NAO etc etc so you're going to be able to produce all these plots this afternoon and then those you know that is kind of you know some examples for the seasonal scale but then you can do something similar too for the subsistional one and this is something like what Casu shows in that 2008 paper so what you are seeing there in those colors is actually the anomalous percentage of occurrence so it's like a measure of frequency and you can imagine that that is a percentage and what you have is in this case those five weather types that we have been using in this particular example and you can say what you have on the vertical axis is the lag so you can say something about when the MJO is in this particular phase let's say phase number six from 14 days to 10 days in advance I know that there are higher chances that I have this particular weather type happening you know after those 14 or 10 days so you can link the same thing that we did before with the seasonal drivers you can just identify in this case I'm just showing MJO but something that we did for Southeast in South America was to consider three different subsistional drivers MJO SAX and then some South American Monsun index which has like a strong peak subsistional scale and you can identify you can try to use this kind of tools to identify which are candidate predictors and how they might or they might not interact at different time scales and if they are associated with tropical or so tropical sources of predictability so I don't know if you know you want me to talk more about this basically blue means that the frequency of occurrence is let's say that blue means less frequent and red means more frequent occurrence of the particular weather type associated with each one of the MJO phases and what you have here is just as I said before the lag the lead time that you are observing and again you can use this kind of plots to diagnose if those links, if those mechanisms appear are well represented or not in your model so I just want to use the rest of my time just to talk a bit about this idea of trying to take advantage of the tropical-exotropical interactions that we are discussing in this advanced school in terms of predictability and how that actually relates to this weather typing method that we have been discussing here and I think that probably you know that figure is just a summary of a qualitative way of what is the forecast skill associated with the weather forecast the seasonal forecast and the subsistional forecast and as we know, I think that we know pretty well, this is a school focused on the intracisional scale so the intracisional scale, our model for the subsistional scale are not that good yet, it depends on when it depends on, you know, if you are considered week 3, week 4 it depends on where but what if, using these tropical-exotropical interactions that we are discussing in this school you can actually pump predictability from different time scales so you can have higher predictive scale for the subsistional for week 2 or week 3 or week 4 together and this has been discussed by several authors and this is just a pictorial representation of that idea so, you know, they talk about the interactions between ENSO and the MEGO and if they are in a constructive phase so they can like reinforce and you might have which is actually the answer with extreme events and we said well ENSO is not the only driver explaining what's happening here because if it's only for El Nino I should have like a lot of rainfall extreme events for southeast in South America but I'm also having a lot of dry events so those interactions between the tropical and the exotropical climate drivers or only, you know, exotropical ones or, you know, high scale or the same one are important to understand what's going on in terms of impacts for society and also in terms of, you know, maybe more academic questions and I don't think I need to show this one but, you know, we can take a look at what is the representation what is the translation of all that in terms of the space phase but, you know, putting all the pieces together what we want to be able to consider our dynamical models should be able to consider those interactions straightforward but that is not always happening and for a particular region I might be able to use this approach that I just showed to you to identify the different candidate predictors try to understand better what's going on in terms of the different spatial and temporal characteristics that are happening the occurrence or not of, you know, rainfall or heat waves in my region and I might be able to identify a lot of those climate drivers this is just a subset, just an example but a different approach is just to try to identify what are those available or proxies for the available states of the system and then use them to try to understand what are the things to those climate drivers and if I if I'm successful at that approach I might be able to build statistical models in terms of the characteristics temporal characteristics for example of those weather regimes or I can even like use those regimes to bias correct the dynamical output of the model and you know, we have been doing a bit of that with the dynamical models doing like some rectification weather type rectification and that has proven to be actually very useful in terms of, you know providing better products for society at different time scales but the overall idea is that since we have a lot of these interactions that we are experiencing it seems more than adequate to pay attention to those possible face locking between climate drivers acting at different time scales and try to take advantage of that in order to have skill enhancement to have it improving of our skill and we have proven that for example for Southeast and South America the use of that approach like considering in our models those interactions provide better skill for extreme rainfall event and as you can see this is following more of a local approach for the forecast of extreme rainfall events which as I said before is extremely difficult and this approach is actually it looks promising but also it might be useful for providing additional information subsistional time scales and just to finish my talk here I just want to describe this approach that we have been using in different places of the world and that is based on these circulation regimes and again is you can consider it a way to build subsistional scenarios and this is not the first time we are not the first people talking about this kind of scenario these are not climate change scenarios just like subsistional to subsistional more on it all have been also working with this approach and other places of the world using a slightly different approach so in our case he has been using let's say that he has built building those scenarios using a cluster of the different rainfall regimes that you may have in that particular region but since models are better are forecasting or you know we are going to be looking at some of the different types of circulation regimes our method is based on those circulation regimes so we start again with that kind of clay diagram and this is the one associated with December, January and February with the austral summer for this particular region as you saw before in a few of my slides trying to construct simplified approach but you can actually try to create clusters with this 90 day sequences of your clay diagram we are not expecting that to be perfect we are not expecting to be able to say what's going to happen in this particular day within the season but we want to see if we can actually build we can classify we can do some kind of analog and say well typically we have this kind of years or that other kind of years and try to take advantage of that information that we have from the observations and depend how well the model reproduce that and if that's good enough we can take advantage of that to say well this year when we have this particular model and we have these phases of MGO at the very beginning of the season what we might expect is this kind of behavior like more rainfall at the beginning of the year or at the end and that has actually proven to be skillful enough I have doubts about using the word skillful because this is not formally a forecast model but it has been able to provide useful information that potentially could be used for decision makers to characterize what might happen in the next season at a subsistional time scale and well there are different ways to do this and I don't know if I have time to describe in detail the approach but it's already published out there and the main message I don't know if actually in terms of time I don't know if I'm early or late okay so the main message here is let's say that you just woke up so the main message here is we might be able not tool is perfect but we might be able to use weather types a circulation a synoptic scale a circulation pattern approach to understand the physical mechanisms behind things like extreme rainfall events but not necessarily extreme rainfall events because the whole idea is that we are using those states to explain the particular event that is of interest to me even if that is like sunny days and we cannot only try to understand better what are the physical mechanisms but we can also try to identify the indicators behind those physical mechanisms and let's say impactful events and have not a perfect but at least a useful approach to try to understand better the forecast that are being provided by dynamical models and that could provide information not only at a more traditional seasonal scale but also to provide additional information at something like the intracisional evolution of the intracisional characteristics of that season of interest to me well we can talk more about that but that's the idea so in practical terms this translates to more work in the labs so I'm going to be able to provide this code in MATLAB it has almost a modified version maybe it's not the most elegant one but you will have there about 2,000 lines of code that you can modify and play with and all the figures that have been shown in particular in this presentation are going to be available for you if you think that that is useful for your project or just to have fun you can modify and tailor that to different regions of the world and for different seasons even different length of the season if you can see that these weather types as building blocks because they are a daily scale you can try to use that to understand different time scales and those tropical interactions that we have been discussing in this school so that's it